Yeah, I’m frustrated. The title says it all. As I said on previous posts (I am trying to like GitHub Spec-Kit, but), I have been trying to like GitHub Spec-Kit, I keep trying, but it seems it does not like me.
At first I tried to do a one-shot spec, just to see the output. I know very well that one-shot’s are not friends yet with real software engineering, but still I wanted to see the output.
Then I went serious and broke functionality into smaller chunks, at first split in full pages, but then I went even smaller and drafted small specs of functionality.
I have been disciplined enough not to mess with the codebase as I did with the BMAD method testing when I did some refactoring work in the middle of the testing. With spec-kit, I have barely done that.
Yet, the results I am getting are not acceptable. I feel I am not making progress. The use case I’m trying to accomplish should not be that hard. I have a chatbot, I type some information, then with the help of AI, I extract some specific pieces of information, which is supposed to be saved in the database. The result is the complete opposite, no information extracted, no information shown, duplicate chatbot instances, and a waste of time defining the spec, plan and tasks. Even though the BMAD method has its own issues, I feel the accuracy in the implementation of features was way higher. I did not have to babysit the whole thing as closely I am doing it now.
I know the framework is new. It’s open source. It’s a nice approach to solving a common problem. It’s free. But yet I right now feel it was premature to let it out (specially if we think it’s coming from Microsoft/GitHub). I just went to the spec-kit repo in GitHub and the slash commands now have changed. They were now prepended with the word ‘specify.’. The API is obviously not locked and there is no direct way to upgrade the framework.
So yeah, either I’m ignorant and I’m doing something wrong or Spec-kit simply isn’t working for me. Whatever the case is, I’m very close to put this on hold and try something else.
This text by no means tries to diminish the value of this type of open-source projects. I think the general idea is great and the approach taken is interesting. I wish I had the time to contribute to it. I appreciate those who do or that make the extra effort to do it.